[Mesa-users] Massive stars and timestep management

[Debashis Sanyal]

Hi Giovanni,

I’m jumping in here since I’ve worked a bit on such massive stars in the past. 
It depends on what you want to achieve with the models. If you’re trying to push the evolution till the pre-SN phase, in my opinion it’s best to start with a naked helium star (assuming the main-sequence star has lost all its envelope of course) and follow its evolution. Although with a very massive helium star you’ll run in to similar problems.

If you want to study the main sequence evolution of such stars itself, there are open questions on what happens when a star develops a bloated/inflated envelope. The point is that the thermal timescale of the stellar envelope becomes very short (of the order of dynamical timescale), which forces the small time steps that MESA (or any other code) takes. It’s not clear at all whether MLT (or MLT++) is a reasonable choice for modelling convection in these layers. You might look at some papers which addresses this point with 3-D simulations.

To answer your question about the trick, you will have to increase the convective flux artificially in the outer stellar layers to push the evolution further.

Best,
Debashis

> On 12. Nov 2020, at 18:11, GM Mirouh via Mesa-users <mesa-users at lists.mesastar.org> wrote:
> 
> Hello everyone,
> 
> I am trying to evolve massive stars (>300Msun) on the main sequence. I have relaxed my initial model from a lower mass, but I have not changed the other parameters wrt what I do for those lower masses. Also I save my models quite regularly on the MS and then restart with a voluntarily small timestep, and I have turned mass loss off (despite its relevance for those stars, I know).
> 
> I have two issues:
> - it is very slow (on 8 cores, my central h1 decreases by ~3% every hour, while it's much faster for lower mass models). I use MESA 12115 and I have attached my inlists for you to have a look, I don't really recommend running them unless you have too many CPU hours on your hands.
> 
> - it sometimes crashes, with a timestep that drops to very low values and NaNs appearing in mlt_partials(mlt_cv_var,mlt_gradT). Not sure what this means, I have attached an example error I got. I know there is an MLT++ option that would probably be relevant for this kind of bloated, warm stars, but I have been advised to not use it unless I understand the details of it, and I do not.
> 
> Do you have any trick to make the massive-star evolution proceed faster and avoid interruptions? Should I use MLT++ or is there a secret combination of timestep and convergence controls that can work? I am mostly interested in modelling the MS at the moment, and could not find much on the mailing list...
> 
> Thank you for any help you can provide, and take care.
> -- 
> Giovanni Mirouh.
> 
> <error.txt><inlist_common.txt><inlist_massive.txt>_______________________________________________
> mesa-users at lists.mesastar.org
> https://lists.mesastar.org/mailman/listinfo/mesa-users
>